• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		IPv4 Forwarding Information Base: semantics.
8  *
9  * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
10  */
11 
12 #include <linux/uaccess.h>
13 #include <linux/bitops.h>
14 #include <linux/types.h>
15 #include <linux/kernel.h>
16 #include <linux/jiffies.h>
17 #include <linux/mm.h>
18 #include <linux/string.h>
19 #include <linux/socket.h>
20 #include <linux/sockios.h>
21 #include <linux/errno.h>
22 #include <linux/in.h>
23 #include <linux/inet.h>
24 #include <linux/inetdevice.h>
25 #include <linux/netdevice.h>
26 #include <linux/if_arp.h>
27 #include <linux/proc_fs.h>
28 #include <linux/skbuff.h>
29 #include <linux/init.h>
30 #include <linux/slab.h>
31 #include <linux/netlink.h>
32 #include <linux/hash.h>
33 #include <linux/nospec.h>
34 
35 #include <net/arp.h>
36 #include <net/ip.h>
37 #include <net/protocol.h>
38 #include <net/route.h>
39 #include <net/tcp.h>
40 #include <net/sock.h>
41 #include <net/ip_fib.h>
42 #include <net/ip6_fib.h>
43 #include <net/nexthop.h>
44 #include <net/netlink.h>
45 #include <net/rtnh.h>
46 #include <net/lwtunnel.h>
47 #include <net/fib_notifier.h>
48 #include <net/addrconf.h>
49 
50 #include "fib_lookup.h"
51 
52 static DEFINE_SPINLOCK(fib_info_lock);
53 static struct hlist_head *fib_info_hash;
54 static struct hlist_head *fib_info_laddrhash;
55 static unsigned int fib_info_hash_size;
56 static unsigned int fib_info_cnt;
57 
58 #define DEVINDEX_HASHBITS 8
59 #define DEVINDEX_HASHSIZE (1U << DEVINDEX_HASHBITS)
60 static struct hlist_head fib_info_devhash[DEVINDEX_HASHSIZE];
61 
62 /* for_nexthops and change_nexthops only used when nexthop object
63  * is not set in a fib_info. The logic within can reference fib_nh.
64  */
65 #ifdef CONFIG_IP_ROUTE_MULTIPATH
66 
67 #define for_nexthops(fi) {						\
68 	int nhsel; const struct fib_nh *nh;				\
69 	for (nhsel = 0, nh = (fi)->fib_nh;				\
70 	     nhsel < fib_info_num_path((fi));				\
71 	     nh++, nhsel++)
72 
73 #define change_nexthops(fi) {						\
74 	int nhsel; struct fib_nh *nexthop_nh;				\
75 	for (nhsel = 0,	nexthop_nh = (struct fib_nh *)((fi)->fib_nh);	\
76 	     nhsel < fib_info_num_path((fi));				\
77 	     nexthop_nh++, nhsel++)
78 
79 #else /* CONFIG_IP_ROUTE_MULTIPATH */
80 
81 /* Hope, that gcc will optimize it to get rid of dummy loop */
82 
83 #define for_nexthops(fi) {						\
84 	int nhsel; const struct fib_nh *nh = (fi)->fib_nh;		\
85 	for (nhsel = 0; nhsel < 1; nhsel++)
86 
87 #define change_nexthops(fi) {						\
88 	int nhsel;							\
89 	struct fib_nh *nexthop_nh = (struct fib_nh *)((fi)->fib_nh);	\
90 	for (nhsel = 0; nhsel < 1; nhsel++)
91 
92 #endif /* CONFIG_IP_ROUTE_MULTIPATH */
93 
94 #define endfor_nexthops(fi) }
95 
96 
97 const struct fib_prop fib_props[RTN_MAX + 1] = {
98 	[RTN_UNSPEC] = {
99 		.error	= 0,
100 		.scope	= RT_SCOPE_NOWHERE,
101 	},
102 	[RTN_UNICAST] = {
103 		.error	= 0,
104 		.scope	= RT_SCOPE_UNIVERSE,
105 	},
106 	[RTN_LOCAL] = {
107 		.error	= 0,
108 		.scope	= RT_SCOPE_HOST,
109 	},
110 	[RTN_BROADCAST] = {
111 		.error	= 0,
112 		.scope	= RT_SCOPE_LINK,
113 	},
114 	[RTN_ANYCAST] = {
115 		.error	= 0,
116 		.scope	= RT_SCOPE_LINK,
117 	},
118 	[RTN_MULTICAST] = {
119 		.error	= 0,
120 		.scope	= RT_SCOPE_UNIVERSE,
121 	},
122 	[RTN_BLACKHOLE] = {
123 		.error	= -EINVAL,
124 		.scope	= RT_SCOPE_UNIVERSE,
125 	},
126 	[RTN_UNREACHABLE] = {
127 		.error	= -EHOSTUNREACH,
128 		.scope	= RT_SCOPE_UNIVERSE,
129 	},
130 	[RTN_PROHIBIT] = {
131 		.error	= -EACCES,
132 		.scope	= RT_SCOPE_UNIVERSE,
133 	},
134 	[RTN_THROW] = {
135 		.error	= -EAGAIN,
136 		.scope	= RT_SCOPE_UNIVERSE,
137 	},
138 	[RTN_NAT] = {
139 		.error	= -EINVAL,
140 		.scope	= RT_SCOPE_NOWHERE,
141 	},
142 	[RTN_XRESOLVE] = {
143 		.error	= -EINVAL,
144 		.scope	= RT_SCOPE_NOWHERE,
145 	},
146 };
147 
rt_fibinfo_free(struct rtable __rcu ** rtp)148 static void rt_fibinfo_free(struct rtable __rcu **rtp)
149 {
150 	struct rtable *rt = rcu_dereference_protected(*rtp, 1);
151 
152 	if (!rt)
153 		return;
154 
155 	/* Not even needed : RCU_INIT_POINTER(*rtp, NULL);
156 	 * because we waited an RCU grace period before calling
157 	 * free_fib_info_rcu()
158 	 */
159 
160 	dst_dev_put(&rt->dst);
161 	dst_release_immediate(&rt->dst);
162 }
163 
free_nh_exceptions(struct fib_nh_common * nhc)164 static void free_nh_exceptions(struct fib_nh_common *nhc)
165 {
166 	struct fnhe_hash_bucket *hash;
167 	int i;
168 
169 	hash = rcu_dereference_protected(nhc->nhc_exceptions, 1);
170 	if (!hash)
171 		return;
172 	for (i = 0; i < FNHE_HASH_SIZE; i++) {
173 		struct fib_nh_exception *fnhe;
174 
175 		fnhe = rcu_dereference_protected(hash[i].chain, 1);
176 		while (fnhe) {
177 			struct fib_nh_exception *next;
178 
179 			next = rcu_dereference_protected(fnhe->fnhe_next, 1);
180 
181 			rt_fibinfo_free(&fnhe->fnhe_rth_input);
182 			rt_fibinfo_free(&fnhe->fnhe_rth_output);
183 
184 			kfree(fnhe);
185 
186 			fnhe = next;
187 		}
188 	}
189 	kfree(hash);
190 }
191 
rt_fibinfo_free_cpus(struct rtable __rcu * __percpu * rtp)192 static void rt_fibinfo_free_cpus(struct rtable __rcu * __percpu *rtp)
193 {
194 	int cpu;
195 
196 	if (!rtp)
197 		return;
198 
199 	for_each_possible_cpu(cpu) {
200 		struct rtable *rt;
201 
202 		rt = rcu_dereference_protected(*per_cpu_ptr(rtp, cpu), 1);
203 		if (rt) {
204 			dst_dev_put(&rt->dst);
205 			dst_release_immediate(&rt->dst);
206 		}
207 	}
208 	free_percpu(rtp);
209 }
210 
fib_nh_common_release(struct fib_nh_common * nhc)211 void fib_nh_common_release(struct fib_nh_common *nhc)
212 {
213 	if (nhc->nhc_dev)
214 		dev_put(nhc->nhc_dev);
215 
216 	lwtstate_put(nhc->nhc_lwtstate);
217 	rt_fibinfo_free_cpus(nhc->nhc_pcpu_rth_output);
218 	rt_fibinfo_free(&nhc->nhc_rth_input);
219 	free_nh_exceptions(nhc);
220 }
221 EXPORT_SYMBOL_GPL(fib_nh_common_release);
222 
fib_nh_release(struct net * net,struct fib_nh * fib_nh)223 void fib_nh_release(struct net *net, struct fib_nh *fib_nh)
224 {
225 #ifdef CONFIG_IP_ROUTE_CLASSID
226 	if (fib_nh->nh_tclassid)
227 		atomic_dec(&net->ipv4.fib_num_tclassid_users);
228 #endif
229 	fib_nh_common_release(&fib_nh->nh_common);
230 }
231 
232 /* Release a nexthop info record */
free_fib_info_rcu(struct rcu_head * head)233 static void free_fib_info_rcu(struct rcu_head *head)
234 {
235 	struct fib_info *fi = container_of(head, struct fib_info, rcu);
236 
237 	if (fi->nh) {
238 		nexthop_put(fi->nh);
239 	} else {
240 		change_nexthops(fi) {
241 			fib_nh_release(fi->fib_net, nexthop_nh);
242 		} endfor_nexthops(fi);
243 	}
244 
245 	ip_fib_metrics_put(fi->fib_metrics);
246 
247 	kfree(fi);
248 }
249 
free_fib_info(struct fib_info * fi)250 void free_fib_info(struct fib_info *fi)
251 {
252 	if (fi->fib_dead == 0) {
253 		pr_warn("Freeing alive fib_info %p\n", fi);
254 		return;
255 	}
256 
257 	call_rcu(&fi->rcu, free_fib_info_rcu);
258 }
259 EXPORT_SYMBOL_GPL(free_fib_info);
260 
fib_release_info(struct fib_info * fi)261 void fib_release_info(struct fib_info *fi)
262 {
263 	spin_lock_bh(&fib_info_lock);
264 	if (fi && --fi->fib_treeref == 0) {
265 		hlist_del(&fi->fib_hash);
266 
267 		/* Paired with READ_ONCE() in fib_create_info(). */
268 		WRITE_ONCE(fib_info_cnt, fib_info_cnt - 1);
269 
270 		if (fi->fib_prefsrc)
271 			hlist_del(&fi->fib_lhash);
272 		if (fi->nh) {
273 			list_del(&fi->nh_list);
274 		} else {
275 			change_nexthops(fi) {
276 				if (!nexthop_nh->fib_nh_dev)
277 					continue;
278 				hlist_del(&nexthop_nh->nh_hash);
279 			} endfor_nexthops(fi)
280 		}
281 		fi->fib_dead = 1;
282 		fib_info_put(fi);
283 	}
284 	spin_unlock_bh(&fib_info_lock);
285 }
286 
nh_comp(struct fib_info * fi,struct fib_info * ofi)287 static inline int nh_comp(struct fib_info *fi, struct fib_info *ofi)
288 {
289 	const struct fib_nh *onh;
290 
291 	if (fi->nh || ofi->nh)
292 		return nexthop_cmp(fi->nh, ofi->nh) ? 0 : -1;
293 
294 	if (ofi->fib_nhs == 0)
295 		return 0;
296 
297 	for_nexthops(fi) {
298 		onh = fib_info_nh(ofi, nhsel);
299 
300 		if (nh->fib_nh_oif != onh->fib_nh_oif ||
301 		    nh->fib_nh_gw_family != onh->fib_nh_gw_family ||
302 		    nh->fib_nh_scope != onh->fib_nh_scope ||
303 #ifdef CONFIG_IP_ROUTE_MULTIPATH
304 		    nh->fib_nh_weight != onh->fib_nh_weight ||
305 #endif
306 #ifdef CONFIG_IP_ROUTE_CLASSID
307 		    nh->nh_tclassid != onh->nh_tclassid ||
308 #endif
309 		    lwtunnel_cmp_encap(nh->fib_nh_lws, onh->fib_nh_lws) ||
310 		    ((nh->fib_nh_flags ^ onh->fib_nh_flags) & ~RTNH_COMPARE_MASK))
311 			return -1;
312 
313 		if (nh->fib_nh_gw_family == AF_INET &&
314 		    nh->fib_nh_gw4 != onh->fib_nh_gw4)
315 			return -1;
316 
317 		if (nh->fib_nh_gw_family == AF_INET6 &&
318 		    ipv6_addr_cmp(&nh->fib_nh_gw6, &onh->fib_nh_gw6))
319 			return -1;
320 	} endfor_nexthops(fi);
321 	return 0;
322 }
323 
fib_devindex_hashfn(unsigned int val)324 static inline unsigned int fib_devindex_hashfn(unsigned int val)
325 {
326 	return hash_32(val, DEVINDEX_HASHBITS);
327 }
328 
329 static struct hlist_head *
fib_info_devhash_bucket(const struct net_device * dev)330 fib_info_devhash_bucket(const struct net_device *dev)
331 {
332 	u32 val = net_hash_mix(dev_net(dev)) ^ dev->ifindex;
333 
334 	return &fib_info_devhash[fib_devindex_hashfn(val)];
335 }
336 
fib_info_hashfn_1(int init_val,u8 protocol,u8 scope,u32 prefsrc,u32 priority)337 static unsigned int fib_info_hashfn_1(int init_val, u8 protocol, u8 scope,
338 				      u32 prefsrc, u32 priority)
339 {
340 	unsigned int val = init_val;
341 
342 	val ^= (protocol << 8) | scope;
343 	val ^= prefsrc;
344 	val ^= priority;
345 
346 	return val;
347 }
348 
fib_info_hashfn_result(unsigned int val)349 static unsigned int fib_info_hashfn_result(unsigned int val)
350 {
351 	unsigned int mask = (fib_info_hash_size - 1);
352 
353 	return (val ^ (val >> 7) ^ (val >> 12)) & mask;
354 }
355 
fib_info_hashfn(struct fib_info * fi)356 static inline unsigned int fib_info_hashfn(struct fib_info *fi)
357 {
358 	unsigned int val;
359 
360 	val = fib_info_hashfn_1(fi->fib_nhs, fi->fib_protocol,
361 				fi->fib_scope, (__force u32)fi->fib_prefsrc,
362 				fi->fib_priority);
363 
364 	if (fi->nh) {
365 		val ^= fib_devindex_hashfn(fi->nh->id);
366 	} else {
367 		for_nexthops(fi) {
368 			val ^= fib_devindex_hashfn(nh->fib_nh_oif);
369 		} endfor_nexthops(fi)
370 	}
371 
372 	return fib_info_hashfn_result(val);
373 }
374 
375 /* no metrics, only nexthop id */
fib_find_info_nh(struct net * net,const struct fib_config * cfg)376 static struct fib_info *fib_find_info_nh(struct net *net,
377 					 const struct fib_config *cfg)
378 {
379 	struct hlist_head *head;
380 	struct fib_info *fi;
381 	unsigned int hash;
382 
383 	hash = fib_info_hashfn_1(fib_devindex_hashfn(cfg->fc_nh_id),
384 				 cfg->fc_protocol, cfg->fc_scope,
385 				 (__force u32)cfg->fc_prefsrc,
386 				 cfg->fc_priority);
387 	hash = fib_info_hashfn_result(hash);
388 	head = &fib_info_hash[hash];
389 
390 	hlist_for_each_entry(fi, head, fib_hash) {
391 		if (!net_eq(fi->fib_net, net))
392 			continue;
393 		if (!fi->nh || fi->nh->id != cfg->fc_nh_id)
394 			continue;
395 		if (cfg->fc_protocol == fi->fib_protocol &&
396 		    cfg->fc_scope == fi->fib_scope &&
397 		    cfg->fc_prefsrc == fi->fib_prefsrc &&
398 		    cfg->fc_priority == fi->fib_priority &&
399 		    cfg->fc_type == fi->fib_type &&
400 		    cfg->fc_table == fi->fib_tb_id &&
401 		    !((cfg->fc_flags ^ fi->fib_flags) & ~RTNH_COMPARE_MASK))
402 			return fi;
403 	}
404 
405 	return NULL;
406 }
407 
fib_find_info(struct fib_info * nfi)408 static struct fib_info *fib_find_info(struct fib_info *nfi)
409 {
410 	struct hlist_head *head;
411 	struct fib_info *fi;
412 	unsigned int hash;
413 
414 	hash = fib_info_hashfn(nfi);
415 	head = &fib_info_hash[hash];
416 
417 	hlist_for_each_entry(fi, head, fib_hash) {
418 		if (!net_eq(fi->fib_net, nfi->fib_net))
419 			continue;
420 		if (fi->fib_nhs != nfi->fib_nhs)
421 			continue;
422 		if (nfi->fib_protocol == fi->fib_protocol &&
423 		    nfi->fib_scope == fi->fib_scope &&
424 		    nfi->fib_prefsrc == fi->fib_prefsrc &&
425 		    nfi->fib_priority == fi->fib_priority &&
426 		    nfi->fib_type == fi->fib_type &&
427 		    nfi->fib_tb_id == fi->fib_tb_id &&
428 		    memcmp(nfi->fib_metrics, fi->fib_metrics,
429 			   sizeof(u32) * RTAX_MAX) == 0 &&
430 		    !((nfi->fib_flags ^ fi->fib_flags) & ~RTNH_COMPARE_MASK) &&
431 		    nh_comp(fi, nfi) == 0)
432 			return fi;
433 	}
434 
435 	return NULL;
436 }
437 
438 /* Check, that the gateway is already configured.
439  * Used only by redirect accept routine.
440  */
ip_fib_check_default(__be32 gw,struct net_device * dev)441 int ip_fib_check_default(__be32 gw, struct net_device *dev)
442 {
443 	struct hlist_head *head;
444 	struct fib_nh *nh;
445 
446 	spin_lock(&fib_info_lock);
447 
448 	head = fib_info_devhash_bucket(dev);
449 
450 	hlist_for_each_entry(nh, head, nh_hash) {
451 		if (nh->fib_nh_dev == dev &&
452 		    nh->fib_nh_gw4 == gw &&
453 		    !(nh->fib_nh_flags & RTNH_F_DEAD)) {
454 			spin_unlock(&fib_info_lock);
455 			return 0;
456 		}
457 	}
458 
459 	spin_unlock(&fib_info_lock);
460 
461 	return -1;
462 }
463 
fib_nlmsg_size(struct fib_info * fi)464 static inline size_t fib_nlmsg_size(struct fib_info *fi)
465 {
466 	size_t payload = NLMSG_ALIGN(sizeof(struct rtmsg))
467 			 + nla_total_size(4) /* RTA_TABLE */
468 			 + nla_total_size(4) /* RTA_DST */
469 			 + nla_total_size(4) /* RTA_PRIORITY */
470 			 + nla_total_size(4) /* RTA_PREFSRC */
471 			 + nla_total_size(TCP_CA_NAME_MAX); /* RTAX_CC_ALGO */
472 	unsigned int nhs = fib_info_num_path(fi);
473 
474 	/* space for nested metrics */
475 	payload += nla_total_size((RTAX_MAX * nla_total_size(4)));
476 
477 	if (fi->nh)
478 		payload += nla_total_size(4); /* RTA_NH_ID */
479 
480 	if (nhs) {
481 		size_t nh_encapsize = 0;
482 		/* Also handles the special case nhs == 1 */
483 
484 		/* each nexthop is packed in an attribute */
485 		size_t nhsize = nla_total_size(sizeof(struct rtnexthop));
486 		unsigned int i;
487 
488 		/* may contain flow and gateway attribute */
489 		nhsize += 2 * nla_total_size(4);
490 
491 		/* grab encap info */
492 		for (i = 0; i < fib_info_num_path(fi); i++) {
493 			struct fib_nh_common *nhc = fib_info_nhc(fi, i);
494 
495 			if (nhc->nhc_lwtstate) {
496 				/* RTA_ENCAP_TYPE */
497 				nh_encapsize += lwtunnel_get_encap_size(
498 						nhc->nhc_lwtstate);
499 				/* RTA_ENCAP */
500 				nh_encapsize +=  nla_total_size(2);
501 			}
502 		}
503 
504 		/* all nexthops are packed in a nested attribute */
505 		payload += nla_total_size((nhs * nhsize) + nh_encapsize);
506 
507 	}
508 
509 	return payload;
510 }
511 
rtmsg_fib(int event,__be32 key,struct fib_alias * fa,int dst_len,u32 tb_id,const struct nl_info * info,unsigned int nlm_flags)512 void rtmsg_fib(int event, __be32 key, struct fib_alias *fa,
513 	       int dst_len, u32 tb_id, const struct nl_info *info,
514 	       unsigned int nlm_flags)
515 {
516 	struct fib_rt_info fri;
517 	struct sk_buff *skb;
518 	u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
519 	int err = -ENOBUFS;
520 
521 	skb = nlmsg_new(fib_nlmsg_size(fa->fa_info), GFP_KERNEL);
522 	if (!skb)
523 		goto errout;
524 
525 	fri.fi = fa->fa_info;
526 	fri.tb_id = tb_id;
527 	fri.dst = key;
528 	fri.dst_len = dst_len;
529 	fri.tos = fa->fa_tos;
530 	fri.type = fa->fa_type;
531 	fri.offload = fa->offload;
532 	fri.trap = fa->trap;
533 	err = fib_dump_info(skb, info->portid, seq, event, &fri, nlm_flags);
534 	if (err < 0) {
535 		/* -EMSGSIZE implies BUG in fib_nlmsg_size() */
536 		WARN_ON(err == -EMSGSIZE);
537 		kfree_skb(skb);
538 		goto errout;
539 	}
540 	rtnl_notify(skb, info->nl_net, info->portid, RTNLGRP_IPV4_ROUTE,
541 		    info->nlh, GFP_KERNEL);
542 	return;
543 errout:
544 	if (err < 0)
545 		rtnl_set_sk_err(info->nl_net, RTNLGRP_IPV4_ROUTE, err);
546 }
547 
fib_detect_death(struct fib_info * fi,int order,struct fib_info ** last_resort,int * last_idx,int dflt)548 static int fib_detect_death(struct fib_info *fi, int order,
549 			    struct fib_info **last_resort, int *last_idx,
550 			    int dflt)
551 {
552 	const struct fib_nh_common *nhc = fib_info_nhc(fi, 0);
553 	struct neighbour *n;
554 	int state = NUD_NONE;
555 
556 	if (likely(nhc->nhc_gw_family == AF_INET))
557 		n = neigh_lookup(&arp_tbl, &nhc->nhc_gw.ipv4, nhc->nhc_dev);
558 	else if (nhc->nhc_gw_family == AF_INET6)
559 		n = neigh_lookup(ipv6_stub->nd_tbl, &nhc->nhc_gw.ipv6,
560 				 nhc->nhc_dev);
561 	else
562 		n = NULL;
563 
564 	if (n) {
565 		state = n->nud_state;
566 		neigh_release(n);
567 	} else {
568 		return 0;
569 	}
570 	if (state == NUD_REACHABLE)
571 		return 0;
572 	if ((state & NUD_VALID) && order != dflt)
573 		return 0;
574 	if ((state & NUD_VALID) ||
575 	    (*last_idx < 0 && order > dflt && state != NUD_INCOMPLETE)) {
576 		*last_resort = fi;
577 		*last_idx = order;
578 	}
579 	return 1;
580 }
581 
fib_nh_common_init(struct net * net,struct fib_nh_common * nhc,struct nlattr * encap,u16 encap_type,void * cfg,gfp_t gfp_flags,struct netlink_ext_ack * extack)582 int fib_nh_common_init(struct net *net, struct fib_nh_common *nhc,
583 		       struct nlattr *encap, u16 encap_type,
584 		       void *cfg, gfp_t gfp_flags,
585 		       struct netlink_ext_ack *extack)
586 {
587 	int err;
588 
589 	nhc->nhc_pcpu_rth_output = alloc_percpu_gfp(struct rtable __rcu *,
590 						    gfp_flags);
591 	if (!nhc->nhc_pcpu_rth_output)
592 		return -ENOMEM;
593 
594 	if (encap) {
595 		struct lwtunnel_state *lwtstate;
596 
597 		if (encap_type == LWTUNNEL_ENCAP_NONE) {
598 			NL_SET_ERR_MSG(extack, "LWT encap type not specified");
599 			err = -EINVAL;
600 			goto lwt_failure;
601 		}
602 		err = lwtunnel_build_state(net, encap_type, encap,
603 					   nhc->nhc_family, cfg, &lwtstate,
604 					   extack);
605 		if (err)
606 			goto lwt_failure;
607 
608 		nhc->nhc_lwtstate = lwtstate_get(lwtstate);
609 	}
610 
611 	return 0;
612 
613 lwt_failure:
614 	rt_fibinfo_free_cpus(nhc->nhc_pcpu_rth_output);
615 	nhc->nhc_pcpu_rth_output = NULL;
616 	return err;
617 }
618 EXPORT_SYMBOL_GPL(fib_nh_common_init);
619 
fib_nh_init(struct net * net,struct fib_nh * nh,struct fib_config * cfg,int nh_weight,struct netlink_ext_ack * extack)620 int fib_nh_init(struct net *net, struct fib_nh *nh,
621 		struct fib_config *cfg, int nh_weight,
622 		struct netlink_ext_ack *extack)
623 {
624 	int err;
625 
626 	nh->fib_nh_family = AF_INET;
627 
628 	err = fib_nh_common_init(net, &nh->nh_common, cfg->fc_encap,
629 				 cfg->fc_encap_type, cfg, GFP_KERNEL, extack);
630 	if (err)
631 		return err;
632 
633 	nh->fib_nh_oif = cfg->fc_oif;
634 	nh->fib_nh_gw_family = cfg->fc_gw_family;
635 	if (cfg->fc_gw_family == AF_INET)
636 		nh->fib_nh_gw4 = cfg->fc_gw4;
637 	else if (cfg->fc_gw_family == AF_INET6)
638 		nh->fib_nh_gw6 = cfg->fc_gw6;
639 
640 	nh->fib_nh_flags = cfg->fc_flags;
641 
642 #ifdef CONFIG_IP_ROUTE_CLASSID
643 	nh->nh_tclassid = cfg->fc_flow;
644 	if (nh->nh_tclassid)
645 		atomic_inc(&net->ipv4.fib_num_tclassid_users);
646 #endif
647 #ifdef CONFIG_IP_ROUTE_MULTIPATH
648 	nh->fib_nh_weight = nh_weight;
649 #endif
650 	return 0;
651 }
652 
653 #ifdef CONFIG_IP_ROUTE_MULTIPATH
654 
fib_count_nexthops(struct rtnexthop * rtnh,int remaining,struct netlink_ext_ack * extack)655 static int fib_count_nexthops(struct rtnexthop *rtnh, int remaining,
656 			      struct netlink_ext_ack *extack)
657 {
658 	int nhs = 0;
659 
660 	while (rtnh_ok(rtnh, remaining)) {
661 		nhs++;
662 		rtnh = rtnh_next(rtnh, &remaining);
663 	}
664 
665 	/* leftover implies invalid nexthop configuration, discard it */
666 	if (remaining > 0) {
667 		NL_SET_ERR_MSG(extack,
668 			       "Invalid nexthop configuration - extra data after nexthops");
669 		nhs = 0;
670 	}
671 
672 	return nhs;
673 }
674 
fib_gw_from_attr(__be32 * gw,struct nlattr * nla,struct netlink_ext_ack * extack)675 static int fib_gw_from_attr(__be32 *gw, struct nlattr *nla,
676 			    struct netlink_ext_ack *extack)
677 {
678 	if (nla_len(nla) < sizeof(*gw)) {
679 		NL_SET_ERR_MSG(extack, "Invalid IPv4 address in RTA_GATEWAY");
680 		return -EINVAL;
681 	}
682 
683 	*gw = nla_get_in_addr(nla);
684 
685 	return 0;
686 }
687 
688 /* only called when fib_nh is integrated into fib_info */
fib_get_nhs(struct fib_info * fi,struct rtnexthop * rtnh,int remaining,struct fib_config * cfg,struct netlink_ext_ack * extack)689 static int fib_get_nhs(struct fib_info *fi, struct rtnexthop *rtnh,
690 		       int remaining, struct fib_config *cfg,
691 		       struct netlink_ext_ack *extack)
692 {
693 	struct net *net = fi->fib_net;
694 	struct fib_config fib_cfg;
695 	struct fib_nh *nh;
696 	int ret;
697 
698 	change_nexthops(fi) {
699 		int attrlen;
700 
701 		memset(&fib_cfg, 0, sizeof(fib_cfg));
702 
703 		if (!rtnh_ok(rtnh, remaining)) {
704 			NL_SET_ERR_MSG(extack,
705 				       "Invalid nexthop configuration - extra data after nexthop");
706 			return -EINVAL;
707 		}
708 
709 		if (rtnh->rtnh_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) {
710 			NL_SET_ERR_MSG(extack,
711 				       "Invalid flags for nexthop - can not contain DEAD or LINKDOWN");
712 			return -EINVAL;
713 		}
714 
715 		fib_cfg.fc_flags = (cfg->fc_flags & ~0xFF) | rtnh->rtnh_flags;
716 		fib_cfg.fc_oif = rtnh->rtnh_ifindex;
717 
718 		attrlen = rtnh_attrlen(rtnh);
719 		if (attrlen > 0) {
720 			struct nlattr *nla, *nlav, *attrs = rtnh_attrs(rtnh);
721 
722 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
723 			nlav = nla_find(attrs, attrlen, RTA_VIA);
724 			if (nla && nlav) {
725 				NL_SET_ERR_MSG(extack,
726 					       "Nexthop configuration can not contain both GATEWAY and VIA");
727 				return -EINVAL;
728 			}
729 			if (nla) {
730 				ret = fib_gw_from_attr(&fib_cfg.fc_gw4, nla,
731 						       extack);
732 				if (ret)
733 					goto errout;
734 
735 				if (fib_cfg.fc_gw4)
736 					fib_cfg.fc_gw_family = AF_INET;
737 			} else if (nlav) {
738 				ret = fib_gw_from_via(&fib_cfg, nlav, extack);
739 				if (ret)
740 					goto errout;
741 			}
742 
743 			nla = nla_find(attrs, attrlen, RTA_FLOW);
744 			if (nla) {
745 				if (nla_len(nla) < sizeof(u32)) {
746 					NL_SET_ERR_MSG(extack, "Invalid RTA_FLOW");
747 					return -EINVAL;
748 				}
749 				fib_cfg.fc_flow = nla_get_u32(nla);
750 			}
751 
752 			fib_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
753 			/* RTA_ENCAP_TYPE length checked in
754 			 * lwtunnel_valid_encap_type_attr
755 			 */
756 			nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
757 			if (nla)
758 				fib_cfg.fc_encap_type = nla_get_u16(nla);
759 		}
760 
761 		ret = fib_nh_init(net, nexthop_nh, &fib_cfg,
762 				  rtnh->rtnh_hops + 1, extack);
763 		if (ret)
764 			goto errout;
765 
766 		rtnh = rtnh_next(rtnh, &remaining);
767 	} endfor_nexthops(fi);
768 
769 	ret = -EINVAL;
770 	nh = fib_info_nh(fi, 0);
771 	if (cfg->fc_oif && nh->fib_nh_oif != cfg->fc_oif) {
772 		NL_SET_ERR_MSG(extack,
773 			       "Nexthop device index does not match RTA_OIF");
774 		goto errout;
775 	}
776 	if (cfg->fc_gw_family) {
777 		if (cfg->fc_gw_family != nh->fib_nh_gw_family ||
778 		    (cfg->fc_gw_family == AF_INET &&
779 		     nh->fib_nh_gw4 != cfg->fc_gw4) ||
780 		    (cfg->fc_gw_family == AF_INET6 &&
781 		     ipv6_addr_cmp(&nh->fib_nh_gw6, &cfg->fc_gw6))) {
782 			NL_SET_ERR_MSG(extack,
783 				       "Nexthop gateway does not match RTA_GATEWAY or RTA_VIA");
784 			goto errout;
785 		}
786 	}
787 #ifdef CONFIG_IP_ROUTE_CLASSID
788 	if (cfg->fc_flow && nh->nh_tclassid != cfg->fc_flow) {
789 		NL_SET_ERR_MSG(extack,
790 			       "Nexthop class id does not match RTA_FLOW");
791 		goto errout;
792 	}
793 #endif
794 	ret = 0;
795 errout:
796 	return ret;
797 }
798 
799 /* only called when fib_nh is integrated into fib_info */
fib_rebalance(struct fib_info * fi)800 static void fib_rebalance(struct fib_info *fi)
801 {
802 	int total;
803 	int w;
804 
805 	if (fib_info_num_path(fi) < 2)
806 		return;
807 
808 	total = 0;
809 	for_nexthops(fi) {
810 		if (nh->fib_nh_flags & RTNH_F_DEAD)
811 			continue;
812 
813 		if (ip_ignore_linkdown(nh->fib_nh_dev) &&
814 		    nh->fib_nh_flags & RTNH_F_LINKDOWN)
815 			continue;
816 
817 		total += nh->fib_nh_weight;
818 	} endfor_nexthops(fi);
819 
820 	w = 0;
821 	change_nexthops(fi) {
822 		int upper_bound;
823 
824 		if (nexthop_nh->fib_nh_flags & RTNH_F_DEAD) {
825 			upper_bound = -1;
826 		} else if (ip_ignore_linkdown(nexthop_nh->fib_nh_dev) &&
827 			   nexthop_nh->fib_nh_flags & RTNH_F_LINKDOWN) {
828 			upper_bound = -1;
829 		} else {
830 			w += nexthop_nh->fib_nh_weight;
831 			upper_bound = DIV_ROUND_CLOSEST_ULL((u64)w << 31,
832 							    total) - 1;
833 		}
834 
835 		atomic_set(&nexthop_nh->fib_nh_upper_bound, upper_bound);
836 	} endfor_nexthops(fi);
837 }
838 #else /* CONFIG_IP_ROUTE_MULTIPATH */
839 
fib_get_nhs(struct fib_info * fi,struct rtnexthop * rtnh,int remaining,struct fib_config * cfg,struct netlink_ext_ack * extack)840 static int fib_get_nhs(struct fib_info *fi, struct rtnexthop *rtnh,
841 		       int remaining, struct fib_config *cfg,
842 		       struct netlink_ext_ack *extack)
843 {
844 	NL_SET_ERR_MSG(extack, "Multipath support not enabled in kernel");
845 
846 	return -EINVAL;
847 }
848 
849 #define fib_rebalance(fi) do { } while (0)
850 
851 #endif /* CONFIG_IP_ROUTE_MULTIPATH */
852 
fib_encap_match(struct net * net,u16 encap_type,struct nlattr * encap,const struct fib_nh * nh,const struct fib_config * cfg,struct netlink_ext_ack * extack)853 static int fib_encap_match(struct net *net, u16 encap_type,
854 			   struct nlattr *encap,
855 			   const struct fib_nh *nh,
856 			   const struct fib_config *cfg,
857 			   struct netlink_ext_ack *extack)
858 {
859 	struct lwtunnel_state *lwtstate;
860 	int ret, result = 0;
861 
862 	if (encap_type == LWTUNNEL_ENCAP_NONE)
863 		return 0;
864 
865 	ret = lwtunnel_build_state(net, encap_type, encap, AF_INET,
866 				   cfg, &lwtstate, extack);
867 	if (!ret) {
868 		result = lwtunnel_cmp_encap(lwtstate, nh->fib_nh_lws);
869 		lwtstate_free(lwtstate);
870 	}
871 
872 	return result;
873 }
874 
fib_nh_match(struct net * net,struct fib_config * cfg,struct fib_info * fi,struct netlink_ext_ack * extack)875 int fib_nh_match(struct net *net, struct fib_config *cfg, struct fib_info *fi,
876 		 struct netlink_ext_ack *extack)
877 {
878 #ifdef CONFIG_IP_ROUTE_MULTIPATH
879 	struct rtnexthop *rtnh;
880 	int remaining;
881 #endif
882 
883 	if (cfg->fc_priority && cfg->fc_priority != fi->fib_priority)
884 		return 1;
885 
886 	if (cfg->fc_nh_id) {
887 		if (fi->nh && cfg->fc_nh_id == fi->nh->id)
888 			return 0;
889 		return 1;
890 	}
891 
892 	if (fi->nh) {
893 		if (cfg->fc_oif || cfg->fc_gw_family || cfg->fc_mp)
894 			return 1;
895 		return 0;
896 	}
897 
898 	if (cfg->fc_oif || cfg->fc_gw_family) {
899 		struct fib_nh *nh;
900 
901 		nh = fib_info_nh(fi, 0);
902 		if (cfg->fc_encap) {
903 			if (fib_encap_match(net, cfg->fc_encap_type,
904 					    cfg->fc_encap, nh, cfg, extack))
905 				return 1;
906 		}
907 #ifdef CONFIG_IP_ROUTE_CLASSID
908 		if (cfg->fc_flow &&
909 		    cfg->fc_flow != nh->nh_tclassid)
910 			return 1;
911 #endif
912 		if ((cfg->fc_oif && cfg->fc_oif != nh->fib_nh_oif) ||
913 		    (cfg->fc_gw_family &&
914 		     cfg->fc_gw_family != nh->fib_nh_gw_family))
915 			return 1;
916 
917 		if (cfg->fc_gw_family == AF_INET &&
918 		    cfg->fc_gw4 != nh->fib_nh_gw4)
919 			return 1;
920 
921 		if (cfg->fc_gw_family == AF_INET6 &&
922 		    ipv6_addr_cmp(&cfg->fc_gw6, &nh->fib_nh_gw6))
923 			return 1;
924 
925 		return 0;
926 	}
927 
928 #ifdef CONFIG_IP_ROUTE_MULTIPATH
929 	if (!cfg->fc_mp)
930 		return 0;
931 
932 	rtnh = cfg->fc_mp;
933 	remaining = cfg->fc_mp_len;
934 
935 	for_nexthops(fi) {
936 		int attrlen;
937 
938 		if (!rtnh_ok(rtnh, remaining))
939 			return -EINVAL;
940 
941 		if (rtnh->rtnh_ifindex && rtnh->rtnh_ifindex != nh->fib_nh_oif)
942 			return 1;
943 
944 		attrlen = rtnh_attrlen(rtnh);
945 		if (attrlen > 0) {
946 			struct nlattr *nla, *nlav, *attrs = rtnh_attrs(rtnh);
947 			int err;
948 
949 			nla = nla_find(attrs, attrlen, RTA_GATEWAY);
950 			nlav = nla_find(attrs, attrlen, RTA_VIA);
951 			if (nla && nlav) {
952 				NL_SET_ERR_MSG(extack,
953 					       "Nexthop configuration can not contain both GATEWAY and VIA");
954 				return -EINVAL;
955 			}
956 
957 			if (nla) {
958 				__be32 gw;
959 
960 				err = fib_gw_from_attr(&gw, nla, extack);
961 				if (err)
962 					return err;
963 
964 				if (nh->fib_nh_gw_family != AF_INET ||
965 				    gw != nh->fib_nh_gw4)
966 					return 1;
967 			} else if (nlav) {
968 				struct fib_config cfg2;
969 
970 				err = fib_gw_from_via(&cfg2, nlav, extack);
971 				if (err)
972 					return err;
973 
974 				switch (nh->fib_nh_gw_family) {
975 				case AF_INET:
976 					if (cfg2.fc_gw_family != AF_INET ||
977 					    cfg2.fc_gw4 != nh->fib_nh_gw4)
978 						return 1;
979 					break;
980 				case AF_INET6:
981 					if (cfg2.fc_gw_family != AF_INET6 ||
982 					    ipv6_addr_cmp(&cfg2.fc_gw6,
983 							  &nh->fib_nh_gw6))
984 						return 1;
985 					break;
986 				}
987 			}
988 
989 #ifdef CONFIG_IP_ROUTE_CLASSID
990 			nla = nla_find(attrs, attrlen, RTA_FLOW);
991 			if (nla) {
992 				if (nla_len(nla) < sizeof(u32)) {
993 					NL_SET_ERR_MSG(extack, "Invalid RTA_FLOW");
994 					return -EINVAL;
995 				}
996 				if (nla_get_u32(nla) != nh->nh_tclassid)
997 					return 1;
998 			}
999 #endif
1000 		}
1001 
1002 		rtnh = rtnh_next(rtnh, &remaining);
1003 	} endfor_nexthops(fi);
1004 #endif
1005 	return 0;
1006 }
1007 
fib_metrics_match(struct fib_config * cfg,struct fib_info * fi)1008 bool fib_metrics_match(struct fib_config *cfg, struct fib_info *fi)
1009 {
1010 	struct nlattr *nla;
1011 	int remaining;
1012 
1013 	if (!cfg->fc_mx)
1014 		return true;
1015 
1016 	nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1017 		int type = nla_type(nla);
1018 		u32 fi_val, val;
1019 
1020 		if (!type)
1021 			continue;
1022 		if (type > RTAX_MAX)
1023 			return false;
1024 
1025 		type = array_index_nospec(type, RTAX_MAX + 1);
1026 		if (type == RTAX_CC_ALGO) {
1027 			char tmp[TCP_CA_NAME_MAX];
1028 			bool ecn_ca = false;
1029 
1030 			nla_strlcpy(tmp, nla, sizeof(tmp));
1031 			val = tcp_ca_get_key_by_name(fi->fib_net, tmp, &ecn_ca);
1032 		} else {
1033 			if (nla_len(nla) != sizeof(u32))
1034 				return false;
1035 			val = nla_get_u32(nla);
1036 		}
1037 
1038 		fi_val = fi->fib_metrics->metrics[type - 1];
1039 		if (type == RTAX_FEATURES)
1040 			fi_val &= ~DST_FEATURE_ECN_CA;
1041 
1042 		if (fi_val != val)
1043 			return false;
1044 	}
1045 
1046 	return true;
1047 }
1048 
fib_check_nh_v6_gw(struct net * net,struct fib_nh * nh,u32 table,struct netlink_ext_ack * extack)1049 static int fib_check_nh_v6_gw(struct net *net, struct fib_nh *nh,
1050 			      u32 table, struct netlink_ext_ack *extack)
1051 {
1052 	struct fib6_config cfg = {
1053 		.fc_table = table,
1054 		.fc_flags = nh->fib_nh_flags | RTF_GATEWAY,
1055 		.fc_ifindex = nh->fib_nh_oif,
1056 		.fc_gateway = nh->fib_nh_gw6,
1057 	};
1058 	struct fib6_nh fib6_nh = {};
1059 	int err;
1060 
1061 	err = ipv6_stub->fib6_nh_init(net, &fib6_nh, &cfg, GFP_KERNEL, extack);
1062 	if (!err) {
1063 		nh->fib_nh_dev = fib6_nh.fib_nh_dev;
1064 		dev_hold(nh->fib_nh_dev);
1065 		nh->fib_nh_oif = nh->fib_nh_dev->ifindex;
1066 		nh->fib_nh_scope = RT_SCOPE_LINK;
1067 
1068 		ipv6_stub->fib6_nh_release(&fib6_nh);
1069 	}
1070 
1071 	return err;
1072 }
1073 
1074 /*
1075  * Picture
1076  * -------
1077  *
1078  * Semantics of nexthop is very messy by historical reasons.
1079  * We have to take into account, that:
1080  * a) gateway can be actually local interface address,
1081  *    so that gatewayed route is direct.
1082  * b) gateway must be on-link address, possibly
1083  *    described not by an ifaddr, but also by a direct route.
1084  * c) If both gateway and interface are specified, they should not
1085  *    contradict.
1086  * d) If we use tunnel routes, gateway could be not on-link.
1087  *
1088  * Attempt to reconcile all of these (alas, self-contradictory) conditions
1089  * results in pretty ugly and hairy code with obscure logic.
1090  *
1091  * I chose to generalized it instead, so that the size
1092  * of code does not increase practically, but it becomes
1093  * much more general.
1094  * Every prefix is assigned a "scope" value: "host" is local address,
1095  * "link" is direct route,
1096  * [ ... "site" ... "interior" ... ]
1097  * and "universe" is true gateway route with global meaning.
1098  *
1099  * Every prefix refers to a set of "nexthop"s (gw, oif),
1100  * where gw must have narrower scope. This recursion stops
1101  * when gw has LOCAL scope or if "nexthop" is declared ONLINK,
1102  * which means that gw is forced to be on link.
1103  *
1104  * Code is still hairy, but now it is apparently logically
1105  * consistent and very flexible. F.e. as by-product it allows
1106  * to co-exists in peace independent exterior and interior
1107  * routing processes.
1108  *
1109  * Normally it looks as following.
1110  *
1111  * {universe prefix}  -> (gw, oif) [scope link]
1112  *		  |
1113  *		  |-> {link prefix} -> (gw, oif) [scope local]
1114  *					|
1115  *					|-> {local prefix} (terminal node)
1116  */
fib_check_nh_v4_gw(struct net * net,struct fib_nh * nh,u32 table,u8 scope,struct netlink_ext_ack * extack)1117 static int fib_check_nh_v4_gw(struct net *net, struct fib_nh *nh, u32 table,
1118 			      u8 scope, struct netlink_ext_ack *extack)
1119 {
1120 	struct net_device *dev;
1121 	struct fib_result res;
1122 	int err = 0;
1123 
1124 	if (nh->fib_nh_flags & RTNH_F_ONLINK) {
1125 		unsigned int addr_type;
1126 
1127 		if (scope >= RT_SCOPE_LINK) {
1128 			NL_SET_ERR_MSG(extack, "Nexthop has invalid scope");
1129 			return -EINVAL;
1130 		}
1131 		dev = __dev_get_by_index(net, nh->fib_nh_oif);
1132 		if (!dev) {
1133 			NL_SET_ERR_MSG(extack, "Nexthop device required for onlink");
1134 			return -ENODEV;
1135 		}
1136 		if (!(dev->flags & IFF_UP)) {
1137 			NL_SET_ERR_MSG(extack, "Nexthop device is not up");
1138 			return -ENETDOWN;
1139 		}
1140 		addr_type = inet_addr_type_dev_table(net, dev, nh->fib_nh_gw4);
1141 		if (addr_type != RTN_UNICAST) {
1142 			NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1143 			return -EINVAL;
1144 		}
1145 		if (!netif_carrier_ok(dev))
1146 			nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1147 		nh->fib_nh_dev = dev;
1148 		dev_hold(dev);
1149 		nh->fib_nh_scope = RT_SCOPE_LINK;
1150 		return 0;
1151 	}
1152 	rcu_read_lock();
1153 	{
1154 		struct fib_table *tbl = NULL;
1155 		struct flowi4 fl4 = {
1156 			.daddr = nh->fib_nh_gw4,
1157 			.flowi4_scope = scope + 1,
1158 			.flowi4_oif = nh->fib_nh_oif,
1159 			.flowi4_iif = LOOPBACK_IFINDEX,
1160 		};
1161 
1162 		/* It is not necessary, but requires a bit of thinking */
1163 		if (fl4.flowi4_scope < RT_SCOPE_LINK)
1164 			fl4.flowi4_scope = RT_SCOPE_LINK;
1165 
1166 		if (table && table != RT_TABLE_MAIN)
1167 			tbl = fib_get_table(net, table);
1168 
1169 		if (tbl)
1170 			err = fib_table_lookup(tbl, &fl4, &res,
1171 					       FIB_LOOKUP_IGNORE_LINKSTATE |
1172 					       FIB_LOOKUP_NOREF);
1173 
1174 		/* on error or if no table given do full lookup. This
1175 		 * is needed for example when nexthops are in the local
1176 		 * table rather than the given table
1177 		 */
1178 		if (!tbl || err) {
1179 			err = fib_lookup(net, &fl4, &res,
1180 					 FIB_LOOKUP_IGNORE_LINKSTATE);
1181 		}
1182 
1183 		if (err) {
1184 			NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1185 			goto out;
1186 		}
1187 	}
1188 
1189 	err = -EINVAL;
1190 	if (res.type != RTN_UNICAST && res.type != RTN_LOCAL) {
1191 		NL_SET_ERR_MSG(extack, "Nexthop has invalid gateway");
1192 		goto out;
1193 	}
1194 	nh->fib_nh_scope = res.scope;
1195 	nh->fib_nh_oif = FIB_RES_OIF(res);
1196 	nh->fib_nh_dev = dev = FIB_RES_DEV(res);
1197 	if (!dev) {
1198 		NL_SET_ERR_MSG(extack,
1199 			       "No egress device for nexthop gateway");
1200 		goto out;
1201 	}
1202 	dev_hold(dev);
1203 	if (!netif_carrier_ok(dev))
1204 		nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1205 	err = (dev->flags & IFF_UP) ? 0 : -ENETDOWN;
1206 out:
1207 	rcu_read_unlock();
1208 	return err;
1209 }
1210 
fib_check_nh_nongw(struct net * net,struct fib_nh * nh,struct netlink_ext_ack * extack)1211 static int fib_check_nh_nongw(struct net *net, struct fib_nh *nh,
1212 			      struct netlink_ext_ack *extack)
1213 {
1214 	struct in_device *in_dev;
1215 	int err;
1216 
1217 	if (nh->fib_nh_flags & (RTNH_F_PERVASIVE | RTNH_F_ONLINK)) {
1218 		NL_SET_ERR_MSG(extack,
1219 			       "Invalid flags for nexthop - PERVASIVE and ONLINK can not be set");
1220 		return -EINVAL;
1221 	}
1222 
1223 	rcu_read_lock();
1224 
1225 	err = -ENODEV;
1226 	in_dev = inetdev_by_index(net, nh->fib_nh_oif);
1227 	if (!in_dev)
1228 		goto out;
1229 	err = -ENETDOWN;
1230 	if (!(in_dev->dev->flags & IFF_UP)) {
1231 		NL_SET_ERR_MSG(extack, "Device for nexthop is not up");
1232 		goto out;
1233 	}
1234 
1235 	nh->fib_nh_dev = in_dev->dev;
1236 	dev_hold(nh->fib_nh_dev);
1237 	nh->fib_nh_scope = RT_SCOPE_LINK;
1238 	if (!netif_carrier_ok(nh->fib_nh_dev))
1239 		nh->fib_nh_flags |= RTNH_F_LINKDOWN;
1240 	err = 0;
1241 out:
1242 	rcu_read_unlock();
1243 	return err;
1244 }
1245 
fib_check_nh(struct net * net,struct fib_nh * nh,u32 table,u8 scope,struct netlink_ext_ack * extack)1246 int fib_check_nh(struct net *net, struct fib_nh *nh, u32 table, u8 scope,
1247 		 struct netlink_ext_ack *extack)
1248 {
1249 	int err;
1250 
1251 	if (nh->fib_nh_gw_family == AF_INET)
1252 		err = fib_check_nh_v4_gw(net, nh, table, scope, extack);
1253 	else if (nh->fib_nh_gw_family == AF_INET6)
1254 		err = fib_check_nh_v6_gw(net, nh, table, extack);
1255 	else
1256 		err = fib_check_nh_nongw(net, nh, extack);
1257 
1258 	return err;
1259 }
1260 
fib_laddr_hashfn(__be32 val)1261 static inline unsigned int fib_laddr_hashfn(__be32 val)
1262 {
1263 	unsigned int mask = (fib_info_hash_size - 1);
1264 
1265 	return ((__force u32)val ^
1266 		((__force u32)val >> 7) ^
1267 		((__force u32)val >> 14)) & mask;
1268 }
1269 
fib_info_hash_alloc(int bytes)1270 static struct hlist_head *fib_info_hash_alloc(int bytes)
1271 {
1272 	if (bytes <= PAGE_SIZE)
1273 		return kzalloc(bytes, GFP_KERNEL);
1274 	else
1275 		return (struct hlist_head *)
1276 			__get_free_pages(GFP_KERNEL | __GFP_ZERO,
1277 					 get_order(bytes));
1278 }
1279 
fib_info_hash_free(struct hlist_head * hash,int bytes)1280 static void fib_info_hash_free(struct hlist_head *hash, int bytes)
1281 {
1282 	if (!hash)
1283 		return;
1284 
1285 	if (bytes <= PAGE_SIZE)
1286 		kfree(hash);
1287 	else
1288 		free_pages((unsigned long) hash, get_order(bytes));
1289 }
1290 
fib_info_hash_move(struct hlist_head * new_info_hash,struct hlist_head * new_laddrhash,unsigned int new_size)1291 static void fib_info_hash_move(struct hlist_head *new_info_hash,
1292 			       struct hlist_head *new_laddrhash,
1293 			       unsigned int new_size)
1294 {
1295 	struct hlist_head *old_info_hash, *old_laddrhash;
1296 	unsigned int old_size = fib_info_hash_size;
1297 	unsigned int i, bytes;
1298 
1299 	spin_lock_bh(&fib_info_lock);
1300 	old_info_hash = fib_info_hash;
1301 	old_laddrhash = fib_info_laddrhash;
1302 	fib_info_hash_size = new_size;
1303 
1304 	for (i = 0; i < old_size; i++) {
1305 		struct hlist_head *head = &fib_info_hash[i];
1306 		struct hlist_node *n;
1307 		struct fib_info *fi;
1308 
1309 		hlist_for_each_entry_safe(fi, n, head, fib_hash) {
1310 			struct hlist_head *dest;
1311 			unsigned int new_hash;
1312 
1313 			new_hash = fib_info_hashfn(fi);
1314 			dest = &new_info_hash[new_hash];
1315 			hlist_add_head(&fi->fib_hash, dest);
1316 		}
1317 	}
1318 	fib_info_hash = new_info_hash;
1319 
1320 	for (i = 0; i < old_size; i++) {
1321 		struct hlist_head *lhead = &fib_info_laddrhash[i];
1322 		struct hlist_node *n;
1323 		struct fib_info *fi;
1324 
1325 		hlist_for_each_entry_safe(fi, n, lhead, fib_lhash) {
1326 			struct hlist_head *ldest;
1327 			unsigned int new_hash;
1328 
1329 			new_hash = fib_laddr_hashfn(fi->fib_prefsrc);
1330 			ldest = &new_laddrhash[new_hash];
1331 			hlist_add_head(&fi->fib_lhash, ldest);
1332 		}
1333 	}
1334 	fib_info_laddrhash = new_laddrhash;
1335 
1336 	spin_unlock_bh(&fib_info_lock);
1337 
1338 	bytes = old_size * sizeof(struct hlist_head *);
1339 	fib_info_hash_free(old_info_hash, bytes);
1340 	fib_info_hash_free(old_laddrhash, bytes);
1341 }
1342 
fib_info_update_nhc_saddr(struct net * net,struct fib_nh_common * nhc,unsigned char scope)1343 __be32 fib_info_update_nhc_saddr(struct net *net, struct fib_nh_common *nhc,
1344 				 unsigned char scope)
1345 {
1346 	struct fib_nh *nh;
1347 
1348 	if (nhc->nhc_family != AF_INET)
1349 		return inet_select_addr(nhc->nhc_dev, 0, scope);
1350 
1351 	nh = container_of(nhc, struct fib_nh, nh_common);
1352 	nh->nh_saddr = inet_select_addr(nh->fib_nh_dev, nh->fib_nh_gw4, scope);
1353 	nh->nh_saddr_genid = atomic_read(&net->ipv4.dev_addr_genid);
1354 
1355 	return nh->nh_saddr;
1356 }
1357 
fib_result_prefsrc(struct net * net,struct fib_result * res)1358 __be32 fib_result_prefsrc(struct net *net, struct fib_result *res)
1359 {
1360 	struct fib_nh_common *nhc = res->nhc;
1361 
1362 	if (res->fi->fib_prefsrc)
1363 		return res->fi->fib_prefsrc;
1364 
1365 	if (nhc->nhc_family == AF_INET) {
1366 		struct fib_nh *nh;
1367 
1368 		nh = container_of(nhc, struct fib_nh, nh_common);
1369 		if (nh->nh_saddr_genid == atomic_read(&net->ipv4.dev_addr_genid))
1370 			return nh->nh_saddr;
1371 	}
1372 
1373 	return fib_info_update_nhc_saddr(net, nhc, res->fi->fib_scope);
1374 }
1375 
fib_valid_prefsrc(struct fib_config * cfg,__be32 fib_prefsrc)1376 static bool fib_valid_prefsrc(struct fib_config *cfg, __be32 fib_prefsrc)
1377 {
1378 	if (cfg->fc_type != RTN_LOCAL || !cfg->fc_dst ||
1379 	    fib_prefsrc != cfg->fc_dst) {
1380 		u32 tb_id = cfg->fc_table;
1381 		int rc;
1382 
1383 		if (tb_id == RT_TABLE_MAIN)
1384 			tb_id = RT_TABLE_LOCAL;
1385 
1386 		rc = inet_addr_type_table(cfg->fc_nlinfo.nl_net,
1387 					  fib_prefsrc, tb_id);
1388 
1389 		if (rc != RTN_LOCAL && tb_id != RT_TABLE_LOCAL) {
1390 			rc = inet_addr_type_table(cfg->fc_nlinfo.nl_net,
1391 						  fib_prefsrc, RT_TABLE_LOCAL);
1392 		}
1393 
1394 		if (rc != RTN_LOCAL)
1395 			return false;
1396 	}
1397 	return true;
1398 }
1399 
fib_create_info(struct fib_config * cfg,struct netlink_ext_ack * extack)1400 struct fib_info *fib_create_info(struct fib_config *cfg,
1401 				 struct netlink_ext_ack *extack)
1402 {
1403 	int err;
1404 	struct fib_info *fi = NULL;
1405 	struct nexthop *nh = NULL;
1406 	struct fib_info *ofi;
1407 	int nhs = 1;
1408 	struct net *net = cfg->fc_nlinfo.nl_net;
1409 
1410 	if (cfg->fc_type > RTN_MAX)
1411 		goto err_inval;
1412 
1413 	/* Fast check to catch the most weird cases */
1414 	if (fib_props[cfg->fc_type].scope > cfg->fc_scope) {
1415 		NL_SET_ERR_MSG(extack, "Invalid scope");
1416 		goto err_inval;
1417 	}
1418 
1419 	if (cfg->fc_flags & (RTNH_F_DEAD | RTNH_F_LINKDOWN)) {
1420 		NL_SET_ERR_MSG(extack,
1421 			       "Invalid rtm_flags - can not contain DEAD or LINKDOWN");
1422 		goto err_inval;
1423 	}
1424 
1425 	if (cfg->fc_nh_id) {
1426 		if (!cfg->fc_mx) {
1427 			fi = fib_find_info_nh(net, cfg);
1428 			if (fi) {
1429 				fi->fib_treeref++;
1430 				return fi;
1431 			}
1432 		}
1433 
1434 		nh = nexthop_find_by_id(net, cfg->fc_nh_id);
1435 		if (!nh) {
1436 			NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
1437 			goto err_inval;
1438 		}
1439 		nhs = 0;
1440 	}
1441 
1442 #ifdef CONFIG_IP_ROUTE_MULTIPATH
1443 	if (cfg->fc_mp) {
1444 		nhs = fib_count_nexthops(cfg->fc_mp, cfg->fc_mp_len, extack);
1445 		if (nhs == 0)
1446 			goto err_inval;
1447 	}
1448 #endif
1449 
1450 	err = -ENOBUFS;
1451 
1452 	/* Paired with WRITE_ONCE() in fib_release_info() */
1453 	if (READ_ONCE(fib_info_cnt) >= fib_info_hash_size) {
1454 		unsigned int new_size = fib_info_hash_size << 1;
1455 		struct hlist_head *new_info_hash;
1456 		struct hlist_head *new_laddrhash;
1457 		unsigned int bytes;
1458 
1459 		if (!new_size)
1460 			new_size = 16;
1461 		bytes = new_size * sizeof(struct hlist_head *);
1462 		new_info_hash = fib_info_hash_alloc(bytes);
1463 		new_laddrhash = fib_info_hash_alloc(bytes);
1464 		if (!new_info_hash || !new_laddrhash) {
1465 			fib_info_hash_free(new_info_hash, bytes);
1466 			fib_info_hash_free(new_laddrhash, bytes);
1467 		} else
1468 			fib_info_hash_move(new_info_hash, new_laddrhash, new_size);
1469 
1470 		if (!fib_info_hash_size)
1471 			goto failure;
1472 	}
1473 
1474 	fi = kzalloc(struct_size(fi, fib_nh, nhs), GFP_KERNEL);
1475 	if (!fi)
1476 		goto failure;
1477 	fi->fib_metrics = ip_fib_metrics_init(fi->fib_net, cfg->fc_mx,
1478 					      cfg->fc_mx_len, extack);
1479 	if (IS_ERR(fi->fib_metrics)) {
1480 		err = PTR_ERR(fi->fib_metrics);
1481 		kfree(fi);
1482 		return ERR_PTR(err);
1483 	}
1484 
1485 	fi->fib_net = net;
1486 	fi->fib_protocol = cfg->fc_protocol;
1487 	fi->fib_scope = cfg->fc_scope;
1488 	fi->fib_flags = cfg->fc_flags;
1489 	fi->fib_priority = cfg->fc_priority;
1490 	fi->fib_prefsrc = cfg->fc_prefsrc;
1491 	fi->fib_type = cfg->fc_type;
1492 	fi->fib_tb_id = cfg->fc_table;
1493 
1494 	fi->fib_nhs = nhs;
1495 	if (nh) {
1496 		if (!nexthop_get(nh)) {
1497 			NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
1498 			err = -EINVAL;
1499 		} else {
1500 			err = 0;
1501 			fi->nh = nh;
1502 		}
1503 	} else {
1504 		change_nexthops(fi) {
1505 			nexthop_nh->nh_parent = fi;
1506 		} endfor_nexthops(fi)
1507 
1508 		if (cfg->fc_mp)
1509 			err = fib_get_nhs(fi, cfg->fc_mp, cfg->fc_mp_len, cfg,
1510 					  extack);
1511 		else
1512 			err = fib_nh_init(net, fi->fib_nh, cfg, 1, extack);
1513 	}
1514 
1515 	if (err != 0)
1516 		goto failure;
1517 
1518 	if (fib_props[cfg->fc_type].error) {
1519 		if (cfg->fc_gw_family || cfg->fc_oif || cfg->fc_mp) {
1520 			NL_SET_ERR_MSG(extack,
1521 				       "Gateway, device and multipath can not be specified for this route type");
1522 			goto err_inval;
1523 		}
1524 		goto link_it;
1525 	} else {
1526 		switch (cfg->fc_type) {
1527 		case RTN_UNICAST:
1528 		case RTN_LOCAL:
1529 		case RTN_BROADCAST:
1530 		case RTN_ANYCAST:
1531 		case RTN_MULTICAST:
1532 			break;
1533 		default:
1534 			NL_SET_ERR_MSG(extack, "Invalid route type");
1535 			goto err_inval;
1536 		}
1537 	}
1538 
1539 	if (cfg->fc_scope > RT_SCOPE_HOST) {
1540 		NL_SET_ERR_MSG(extack, "Invalid scope");
1541 		goto err_inval;
1542 	}
1543 
1544 	if (fi->nh) {
1545 		err = fib_check_nexthop(fi->nh, cfg->fc_scope, extack);
1546 		if (err)
1547 			goto failure;
1548 	} else if (cfg->fc_scope == RT_SCOPE_HOST) {
1549 		struct fib_nh *nh = fi->fib_nh;
1550 
1551 		/* Local address is added. */
1552 		if (nhs != 1) {
1553 			NL_SET_ERR_MSG(extack,
1554 				       "Route with host scope can not have multiple nexthops");
1555 			goto err_inval;
1556 		}
1557 		if (nh->fib_nh_gw_family) {
1558 			NL_SET_ERR_MSG(extack,
1559 				       "Route with host scope can not have a gateway");
1560 			goto err_inval;
1561 		}
1562 		nh->fib_nh_scope = RT_SCOPE_NOWHERE;
1563 		nh->fib_nh_dev = dev_get_by_index(net, nh->fib_nh_oif);
1564 		err = -ENODEV;
1565 		if (!nh->fib_nh_dev)
1566 			goto failure;
1567 	} else {
1568 		int linkdown = 0;
1569 
1570 		change_nexthops(fi) {
1571 			err = fib_check_nh(cfg->fc_nlinfo.nl_net, nexthop_nh,
1572 					   cfg->fc_table, cfg->fc_scope,
1573 					   extack);
1574 			if (err != 0)
1575 				goto failure;
1576 			if (nexthop_nh->fib_nh_flags & RTNH_F_LINKDOWN)
1577 				linkdown++;
1578 		} endfor_nexthops(fi)
1579 		if (linkdown == fi->fib_nhs)
1580 			fi->fib_flags |= RTNH_F_LINKDOWN;
1581 	}
1582 
1583 	if (fi->fib_prefsrc && !fib_valid_prefsrc(cfg, fi->fib_prefsrc)) {
1584 		NL_SET_ERR_MSG(extack, "Invalid prefsrc address");
1585 		goto err_inval;
1586 	}
1587 
1588 	if (!fi->nh) {
1589 		change_nexthops(fi) {
1590 			fib_info_update_nhc_saddr(net, &nexthop_nh->nh_common,
1591 						  fi->fib_scope);
1592 			if (nexthop_nh->fib_nh_gw_family == AF_INET6)
1593 				fi->fib_nh_is_v6 = true;
1594 		} endfor_nexthops(fi)
1595 
1596 		fib_rebalance(fi);
1597 	}
1598 
1599 link_it:
1600 	ofi = fib_find_info(fi);
1601 	if (ofi) {
1602 		fi->fib_dead = 1;
1603 		free_fib_info(fi);
1604 		ofi->fib_treeref++;
1605 		return ofi;
1606 	}
1607 
1608 	fi->fib_treeref++;
1609 	refcount_set(&fi->fib_clntref, 1);
1610 	spin_lock_bh(&fib_info_lock);
1611 	fib_info_cnt++;
1612 	hlist_add_head(&fi->fib_hash,
1613 		       &fib_info_hash[fib_info_hashfn(fi)]);
1614 	if (fi->fib_prefsrc) {
1615 		struct hlist_head *head;
1616 
1617 		head = &fib_info_laddrhash[fib_laddr_hashfn(fi->fib_prefsrc)];
1618 		hlist_add_head(&fi->fib_lhash, head);
1619 	}
1620 	if (fi->nh) {
1621 		list_add(&fi->nh_list, &nh->fi_list);
1622 	} else {
1623 		change_nexthops(fi) {
1624 			struct hlist_head *head;
1625 
1626 			if (!nexthop_nh->fib_nh_dev)
1627 				continue;
1628 			head = fib_info_devhash_bucket(nexthop_nh->fib_nh_dev);
1629 			hlist_add_head(&nexthop_nh->nh_hash, head);
1630 		} endfor_nexthops(fi)
1631 	}
1632 	spin_unlock_bh(&fib_info_lock);
1633 	return fi;
1634 
1635 err_inval:
1636 	err = -EINVAL;
1637 
1638 failure:
1639 	if (fi) {
1640 		fi->fib_dead = 1;
1641 		free_fib_info(fi);
1642 	}
1643 
1644 	return ERR_PTR(err);
1645 }
1646 
fib_nexthop_info(struct sk_buff * skb,const struct fib_nh_common * nhc,u8 rt_family,unsigned char * flags,bool skip_oif)1647 int fib_nexthop_info(struct sk_buff *skb, const struct fib_nh_common *nhc,
1648 		     u8 rt_family, unsigned char *flags, bool skip_oif)
1649 {
1650 	if (nhc->nhc_flags & RTNH_F_DEAD)
1651 		*flags |= RTNH_F_DEAD;
1652 
1653 	if (nhc->nhc_flags & RTNH_F_LINKDOWN) {
1654 		*flags |= RTNH_F_LINKDOWN;
1655 
1656 		rcu_read_lock();
1657 		switch (nhc->nhc_family) {
1658 		case AF_INET:
1659 			if (ip_ignore_linkdown(nhc->nhc_dev))
1660 				*flags |= RTNH_F_DEAD;
1661 			break;
1662 		case AF_INET6:
1663 			if (ip6_ignore_linkdown(nhc->nhc_dev))
1664 				*flags |= RTNH_F_DEAD;
1665 			break;
1666 		}
1667 		rcu_read_unlock();
1668 	}
1669 
1670 	switch (nhc->nhc_gw_family) {
1671 	case AF_INET:
1672 		if (nla_put_in_addr(skb, RTA_GATEWAY, nhc->nhc_gw.ipv4))
1673 			goto nla_put_failure;
1674 		break;
1675 	case AF_INET6:
1676 		/* if gateway family does not match nexthop family
1677 		 * gateway is encoded as RTA_VIA
1678 		 */
1679 		if (rt_family != nhc->nhc_gw_family) {
1680 			int alen = sizeof(struct in6_addr);
1681 			struct nlattr *nla;
1682 			struct rtvia *via;
1683 
1684 			nla = nla_reserve(skb, RTA_VIA, alen + 2);
1685 			if (!nla)
1686 				goto nla_put_failure;
1687 
1688 			via = nla_data(nla);
1689 			via->rtvia_family = AF_INET6;
1690 			memcpy(via->rtvia_addr, &nhc->nhc_gw.ipv6, alen);
1691 		} else if (nla_put_in6_addr(skb, RTA_GATEWAY,
1692 					    &nhc->nhc_gw.ipv6) < 0) {
1693 			goto nla_put_failure;
1694 		}
1695 		break;
1696 	}
1697 
1698 	*flags |= (nhc->nhc_flags & RTNH_F_ONLINK);
1699 	if (nhc->nhc_flags & RTNH_F_OFFLOAD)
1700 		*flags |= RTNH_F_OFFLOAD;
1701 
1702 	if (!skip_oif && nhc->nhc_dev &&
1703 	    nla_put_u32(skb, RTA_OIF, nhc->nhc_dev->ifindex))
1704 		goto nla_put_failure;
1705 
1706 	if (nhc->nhc_lwtstate &&
1707 	    lwtunnel_fill_encap(skb, nhc->nhc_lwtstate,
1708 				RTA_ENCAP, RTA_ENCAP_TYPE) < 0)
1709 		goto nla_put_failure;
1710 
1711 	return 0;
1712 
1713 nla_put_failure:
1714 	return -EMSGSIZE;
1715 }
1716 EXPORT_SYMBOL_GPL(fib_nexthop_info);
1717 
1718 #if IS_ENABLED(CONFIG_IP_ROUTE_MULTIPATH) || IS_ENABLED(CONFIG_IPV6)
fib_add_nexthop(struct sk_buff * skb,const struct fib_nh_common * nhc,int nh_weight,u8 rt_family,u32 nh_tclassid)1719 int fib_add_nexthop(struct sk_buff *skb, const struct fib_nh_common *nhc,
1720 		    int nh_weight, u8 rt_family, u32 nh_tclassid)
1721 {
1722 	const struct net_device *dev = nhc->nhc_dev;
1723 	struct rtnexthop *rtnh;
1724 	unsigned char flags = 0;
1725 
1726 	rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
1727 	if (!rtnh)
1728 		goto nla_put_failure;
1729 
1730 	rtnh->rtnh_hops = nh_weight - 1;
1731 	rtnh->rtnh_ifindex = dev ? dev->ifindex : 0;
1732 
1733 	if (fib_nexthop_info(skb, nhc, rt_family, &flags, true) < 0)
1734 		goto nla_put_failure;
1735 
1736 	rtnh->rtnh_flags = flags;
1737 
1738 	if (nh_tclassid && nla_put_u32(skb, RTA_FLOW, nh_tclassid))
1739 		goto nla_put_failure;
1740 
1741 	/* length of rtnetlink header + attributes */
1742 	rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
1743 
1744 	return 0;
1745 
1746 nla_put_failure:
1747 	return -EMSGSIZE;
1748 }
1749 EXPORT_SYMBOL_GPL(fib_add_nexthop);
1750 #endif
1751 
1752 #ifdef CONFIG_IP_ROUTE_MULTIPATH
fib_add_multipath(struct sk_buff * skb,struct fib_info * fi)1753 static int fib_add_multipath(struct sk_buff *skb, struct fib_info *fi)
1754 {
1755 	struct nlattr *mp;
1756 
1757 	mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
1758 	if (!mp)
1759 		goto nla_put_failure;
1760 
1761 	if (unlikely(fi->nh)) {
1762 		if (nexthop_mpath_fill_node(skb, fi->nh, AF_INET) < 0)
1763 			goto nla_put_failure;
1764 		goto mp_end;
1765 	}
1766 
1767 	for_nexthops(fi) {
1768 		u32 nh_tclassid = 0;
1769 #ifdef CONFIG_IP_ROUTE_CLASSID
1770 		nh_tclassid = nh->nh_tclassid;
1771 #endif
1772 		if (fib_add_nexthop(skb, &nh->nh_common, nh->fib_nh_weight,
1773 				    AF_INET, nh_tclassid) < 0)
1774 			goto nla_put_failure;
1775 	} endfor_nexthops(fi);
1776 
1777 mp_end:
1778 	nla_nest_end(skb, mp);
1779 
1780 	return 0;
1781 
1782 nla_put_failure:
1783 	return -EMSGSIZE;
1784 }
1785 #else
fib_add_multipath(struct sk_buff * skb,struct fib_info * fi)1786 static int fib_add_multipath(struct sk_buff *skb, struct fib_info *fi)
1787 {
1788 	return 0;
1789 }
1790 #endif
1791 
fib_dump_info(struct sk_buff * skb,u32 portid,u32 seq,int event,struct fib_rt_info * fri,unsigned int flags)1792 int fib_dump_info(struct sk_buff *skb, u32 portid, u32 seq, int event,
1793 		  struct fib_rt_info *fri, unsigned int flags)
1794 {
1795 	unsigned int nhs = fib_info_num_path(fri->fi);
1796 	struct fib_info *fi = fri->fi;
1797 	u32 tb_id = fri->tb_id;
1798 	struct nlmsghdr *nlh;
1799 	struct rtmsg *rtm;
1800 
1801 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags);
1802 	if (!nlh)
1803 		return -EMSGSIZE;
1804 
1805 	rtm = nlmsg_data(nlh);
1806 	rtm->rtm_family = AF_INET;
1807 	rtm->rtm_dst_len = fri->dst_len;
1808 	rtm->rtm_src_len = 0;
1809 	rtm->rtm_tos = fri->tos;
1810 	if (tb_id < 256)
1811 		rtm->rtm_table = tb_id;
1812 	else
1813 		rtm->rtm_table = RT_TABLE_COMPAT;
1814 	if (nla_put_u32(skb, RTA_TABLE, tb_id))
1815 		goto nla_put_failure;
1816 	rtm->rtm_type = fri->type;
1817 	rtm->rtm_flags = fi->fib_flags;
1818 	rtm->rtm_scope = fi->fib_scope;
1819 	rtm->rtm_protocol = fi->fib_protocol;
1820 
1821 	if (rtm->rtm_dst_len &&
1822 	    nla_put_in_addr(skb, RTA_DST, fri->dst))
1823 		goto nla_put_failure;
1824 	if (fi->fib_priority &&
1825 	    nla_put_u32(skb, RTA_PRIORITY, fi->fib_priority))
1826 		goto nla_put_failure;
1827 	if (rtnetlink_put_metrics(skb, fi->fib_metrics->metrics) < 0)
1828 		goto nla_put_failure;
1829 
1830 	if (fi->fib_prefsrc &&
1831 	    nla_put_in_addr(skb, RTA_PREFSRC, fi->fib_prefsrc))
1832 		goto nla_put_failure;
1833 
1834 	if (fi->nh) {
1835 		if (nla_put_u32(skb, RTA_NH_ID, fi->nh->id))
1836 			goto nla_put_failure;
1837 		if (nexthop_is_blackhole(fi->nh))
1838 			rtm->rtm_type = RTN_BLACKHOLE;
1839 		if (!READ_ONCE(fi->fib_net->ipv4.sysctl_nexthop_compat_mode))
1840 			goto offload;
1841 	}
1842 
1843 	if (nhs == 1) {
1844 		const struct fib_nh_common *nhc = fib_info_nhc(fi, 0);
1845 		unsigned char flags = 0;
1846 
1847 		if (fib_nexthop_info(skb, nhc, AF_INET, &flags, false) < 0)
1848 			goto nla_put_failure;
1849 
1850 		rtm->rtm_flags = flags;
1851 #ifdef CONFIG_IP_ROUTE_CLASSID
1852 		if (nhc->nhc_family == AF_INET) {
1853 			struct fib_nh *nh;
1854 
1855 			nh = container_of(nhc, struct fib_nh, nh_common);
1856 			if (nh->nh_tclassid &&
1857 			    nla_put_u32(skb, RTA_FLOW, nh->nh_tclassid))
1858 				goto nla_put_failure;
1859 		}
1860 #endif
1861 	} else {
1862 		if (fib_add_multipath(skb, fi) < 0)
1863 			goto nla_put_failure;
1864 	}
1865 
1866 offload:
1867 	if (fri->offload)
1868 		rtm->rtm_flags |= RTM_F_OFFLOAD;
1869 	if (fri->trap)
1870 		rtm->rtm_flags |= RTM_F_TRAP;
1871 
1872 	nlmsg_end(skb, nlh);
1873 	return 0;
1874 
1875 nla_put_failure:
1876 	nlmsg_cancel(skb, nlh);
1877 	return -EMSGSIZE;
1878 }
1879 
1880 /*
1881  * Update FIB if:
1882  * - local address disappeared -> we must delete all the entries
1883  *   referring to it.
1884  * - device went down -> we must shutdown all nexthops going via it.
1885  */
fib_sync_down_addr(struct net_device * dev,__be32 local)1886 int fib_sync_down_addr(struct net_device *dev, __be32 local)
1887 {
1888 	int ret = 0;
1889 	unsigned int hash = fib_laddr_hashfn(local);
1890 	struct hlist_head *head = &fib_info_laddrhash[hash];
1891 	int tb_id = l3mdev_fib_table(dev) ? : RT_TABLE_MAIN;
1892 	struct net *net = dev_net(dev);
1893 	struct fib_info *fi;
1894 
1895 	if (!fib_info_laddrhash || local == 0)
1896 		return 0;
1897 
1898 	hlist_for_each_entry(fi, head, fib_lhash) {
1899 		if (!net_eq(fi->fib_net, net) ||
1900 		    fi->fib_tb_id != tb_id)
1901 			continue;
1902 		if (fi->fib_prefsrc == local) {
1903 			fi->fib_flags |= RTNH_F_DEAD;
1904 			ret++;
1905 		}
1906 	}
1907 	return ret;
1908 }
1909 
call_fib_nh_notifiers(struct fib_nh * nh,enum fib_event_type event_type)1910 static int call_fib_nh_notifiers(struct fib_nh *nh,
1911 				 enum fib_event_type event_type)
1912 {
1913 	bool ignore_link_down = ip_ignore_linkdown(nh->fib_nh_dev);
1914 	struct fib_nh_notifier_info info = {
1915 		.fib_nh = nh,
1916 	};
1917 
1918 	switch (event_type) {
1919 	case FIB_EVENT_NH_ADD:
1920 		if (nh->fib_nh_flags & RTNH_F_DEAD)
1921 			break;
1922 		if (ignore_link_down && nh->fib_nh_flags & RTNH_F_LINKDOWN)
1923 			break;
1924 		return call_fib4_notifiers(dev_net(nh->fib_nh_dev), event_type,
1925 					   &info.info);
1926 	case FIB_EVENT_NH_DEL:
1927 		if ((ignore_link_down && nh->fib_nh_flags & RTNH_F_LINKDOWN) ||
1928 		    (nh->fib_nh_flags & RTNH_F_DEAD))
1929 			return call_fib4_notifiers(dev_net(nh->fib_nh_dev),
1930 						   event_type, &info.info);
1931 	default:
1932 		break;
1933 	}
1934 
1935 	return NOTIFY_DONE;
1936 }
1937 
1938 /* Update the PMTU of exceptions when:
1939  * - the new MTU of the first hop becomes smaller than the PMTU
1940  * - the old MTU was the same as the PMTU, and it limited discovery of
1941  *   larger MTUs on the path. With that limit raised, we can now
1942  *   discover larger MTUs
1943  * A special case is locked exceptions, for which the PMTU is smaller
1944  * than the minimal accepted PMTU:
1945  * - if the new MTU is greater than the PMTU, don't make any change
1946  * - otherwise, unlock and set PMTU
1947  */
fib_nhc_update_mtu(struct fib_nh_common * nhc,u32 new,u32 orig)1948 void fib_nhc_update_mtu(struct fib_nh_common *nhc, u32 new, u32 orig)
1949 {
1950 	struct fnhe_hash_bucket *bucket;
1951 	int i;
1952 
1953 	bucket = rcu_dereference_protected(nhc->nhc_exceptions, 1);
1954 	if (!bucket)
1955 		return;
1956 
1957 	for (i = 0; i < FNHE_HASH_SIZE; i++) {
1958 		struct fib_nh_exception *fnhe;
1959 
1960 		for (fnhe = rcu_dereference_protected(bucket[i].chain, 1);
1961 		     fnhe;
1962 		     fnhe = rcu_dereference_protected(fnhe->fnhe_next, 1)) {
1963 			if (fnhe->fnhe_mtu_locked) {
1964 				if (new <= fnhe->fnhe_pmtu) {
1965 					fnhe->fnhe_pmtu = new;
1966 					fnhe->fnhe_mtu_locked = false;
1967 				}
1968 			} else if (new < fnhe->fnhe_pmtu ||
1969 				   orig == fnhe->fnhe_pmtu) {
1970 				fnhe->fnhe_pmtu = new;
1971 			}
1972 		}
1973 	}
1974 }
1975 
fib_sync_mtu(struct net_device * dev,u32 orig_mtu)1976 void fib_sync_mtu(struct net_device *dev, u32 orig_mtu)
1977 {
1978 	struct hlist_head *head = fib_info_devhash_bucket(dev);
1979 	struct fib_nh *nh;
1980 
1981 	hlist_for_each_entry(nh, head, nh_hash) {
1982 		if (nh->fib_nh_dev == dev)
1983 			fib_nhc_update_mtu(&nh->nh_common, dev->mtu, orig_mtu);
1984 	}
1985 }
1986 
1987 /* Event              force Flags           Description
1988  * NETDEV_CHANGE      0     LINKDOWN        Carrier OFF, not for scope host
1989  * NETDEV_DOWN        0     LINKDOWN|DEAD   Link down, not for scope host
1990  * NETDEV_DOWN        1     LINKDOWN|DEAD   Last address removed
1991  * NETDEV_UNREGISTER  1     LINKDOWN|DEAD   Device removed
1992  *
1993  * only used when fib_nh is built into fib_info
1994  */
fib_sync_down_dev(struct net_device * dev,unsigned long event,bool force)1995 int fib_sync_down_dev(struct net_device *dev, unsigned long event, bool force)
1996 {
1997 	struct hlist_head *head = fib_info_devhash_bucket(dev);
1998 	struct fib_info *prev_fi = NULL;
1999 	int scope = RT_SCOPE_NOWHERE;
2000 	struct fib_nh *nh;
2001 	int ret = 0;
2002 
2003 	if (force)
2004 		scope = -1;
2005 
2006 	hlist_for_each_entry(nh, head, nh_hash) {
2007 		struct fib_info *fi = nh->nh_parent;
2008 		int dead;
2009 
2010 		BUG_ON(!fi->fib_nhs);
2011 		if (nh->fib_nh_dev != dev || fi == prev_fi)
2012 			continue;
2013 		prev_fi = fi;
2014 		dead = 0;
2015 		change_nexthops(fi) {
2016 			if (nexthop_nh->fib_nh_flags & RTNH_F_DEAD)
2017 				dead++;
2018 			else if (nexthop_nh->fib_nh_dev == dev &&
2019 				 nexthop_nh->fib_nh_scope != scope) {
2020 				switch (event) {
2021 				case NETDEV_DOWN:
2022 				case NETDEV_UNREGISTER:
2023 					nexthop_nh->fib_nh_flags |= RTNH_F_DEAD;
2024 					fallthrough;
2025 				case NETDEV_CHANGE:
2026 					nexthop_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
2027 					break;
2028 				}
2029 				call_fib_nh_notifiers(nexthop_nh,
2030 						      FIB_EVENT_NH_DEL);
2031 				dead++;
2032 			}
2033 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2034 			if (event == NETDEV_UNREGISTER &&
2035 			    nexthop_nh->fib_nh_dev == dev) {
2036 				dead = fi->fib_nhs;
2037 				break;
2038 			}
2039 #endif
2040 		} endfor_nexthops(fi)
2041 		if (dead == fi->fib_nhs) {
2042 			switch (event) {
2043 			case NETDEV_DOWN:
2044 			case NETDEV_UNREGISTER:
2045 				fi->fib_flags |= RTNH_F_DEAD;
2046 				fallthrough;
2047 			case NETDEV_CHANGE:
2048 				fi->fib_flags |= RTNH_F_LINKDOWN;
2049 				break;
2050 			}
2051 			ret++;
2052 		}
2053 
2054 		fib_rebalance(fi);
2055 	}
2056 
2057 	return ret;
2058 }
2059 
2060 /* Must be invoked inside of an RCU protected region.  */
fib_select_default(const struct flowi4 * flp,struct fib_result * res)2061 static void fib_select_default(const struct flowi4 *flp, struct fib_result *res)
2062 {
2063 	struct fib_info *fi = NULL, *last_resort = NULL;
2064 	struct hlist_head *fa_head = res->fa_head;
2065 	struct fib_table *tb = res->table;
2066 	u8 slen = 32 - res->prefixlen;
2067 	int order = -1, last_idx = -1;
2068 	struct fib_alias *fa, *fa1 = NULL;
2069 	u32 last_prio = res->fi->fib_priority;
2070 	u8 last_tos = 0;
2071 
2072 	hlist_for_each_entry_rcu(fa, fa_head, fa_list) {
2073 		struct fib_info *next_fi = fa->fa_info;
2074 		struct fib_nh_common *nhc;
2075 
2076 		if (fa->fa_slen != slen)
2077 			continue;
2078 		if (fa->fa_tos && fa->fa_tos != flp->flowi4_tos)
2079 			continue;
2080 		if (fa->tb_id != tb->tb_id)
2081 			continue;
2082 		if (next_fi->fib_priority > last_prio &&
2083 		    fa->fa_tos == last_tos) {
2084 			if (last_tos)
2085 				continue;
2086 			break;
2087 		}
2088 		if (next_fi->fib_flags & RTNH_F_DEAD)
2089 			continue;
2090 		last_tos = fa->fa_tos;
2091 		last_prio = next_fi->fib_priority;
2092 
2093 		if (next_fi->fib_scope != res->scope ||
2094 		    fa->fa_type != RTN_UNICAST)
2095 			continue;
2096 
2097 		nhc = fib_info_nhc(next_fi, 0);
2098 		if (!nhc->nhc_gw_family || nhc->nhc_scope != RT_SCOPE_LINK)
2099 			continue;
2100 
2101 		fib_alias_accessed(fa);
2102 
2103 		if (!fi) {
2104 			if (next_fi != res->fi)
2105 				break;
2106 			fa1 = fa;
2107 		} else if (!fib_detect_death(fi, order, &last_resort,
2108 					     &last_idx, fa1->fa_default)) {
2109 			fib_result_assign(res, fi);
2110 			fa1->fa_default = order;
2111 			goto out;
2112 		}
2113 		fi = next_fi;
2114 		order++;
2115 	}
2116 
2117 	if (order <= 0 || !fi) {
2118 		if (fa1)
2119 			fa1->fa_default = -1;
2120 		goto out;
2121 	}
2122 
2123 	if (!fib_detect_death(fi, order, &last_resort, &last_idx,
2124 			      fa1->fa_default)) {
2125 		fib_result_assign(res, fi);
2126 		fa1->fa_default = order;
2127 		goto out;
2128 	}
2129 
2130 	if (last_idx >= 0)
2131 		fib_result_assign(res, last_resort);
2132 	fa1->fa_default = last_idx;
2133 out:
2134 	return;
2135 }
2136 
2137 /*
2138  * Dead device goes up. We wake up dead nexthops.
2139  * It takes sense only on multipath routes.
2140  *
2141  * only used when fib_nh is built into fib_info
2142  */
fib_sync_up(struct net_device * dev,unsigned char nh_flags)2143 int fib_sync_up(struct net_device *dev, unsigned char nh_flags)
2144 {
2145 	struct fib_info *prev_fi;
2146 	struct hlist_head *head;
2147 	struct fib_nh *nh;
2148 	int ret;
2149 
2150 	if (!(dev->flags & IFF_UP))
2151 		return 0;
2152 
2153 	if (nh_flags & RTNH_F_DEAD) {
2154 		unsigned int flags = dev_get_flags(dev);
2155 
2156 		if (flags & (IFF_RUNNING | IFF_LOWER_UP))
2157 			nh_flags |= RTNH_F_LINKDOWN;
2158 	}
2159 
2160 	prev_fi = NULL;
2161 	head = fib_info_devhash_bucket(dev);
2162 	ret = 0;
2163 
2164 	hlist_for_each_entry(nh, head, nh_hash) {
2165 		struct fib_info *fi = nh->nh_parent;
2166 		int alive;
2167 
2168 		BUG_ON(!fi->fib_nhs);
2169 		if (nh->fib_nh_dev != dev || fi == prev_fi)
2170 			continue;
2171 
2172 		prev_fi = fi;
2173 		alive = 0;
2174 		change_nexthops(fi) {
2175 			if (!(nexthop_nh->fib_nh_flags & nh_flags)) {
2176 				alive++;
2177 				continue;
2178 			}
2179 			if (!nexthop_nh->fib_nh_dev ||
2180 			    !(nexthop_nh->fib_nh_dev->flags & IFF_UP))
2181 				continue;
2182 			if (nexthop_nh->fib_nh_dev != dev ||
2183 			    !__in_dev_get_rtnl(dev))
2184 				continue;
2185 			alive++;
2186 			nexthop_nh->fib_nh_flags &= ~nh_flags;
2187 			call_fib_nh_notifiers(nexthop_nh, FIB_EVENT_NH_ADD);
2188 		} endfor_nexthops(fi)
2189 
2190 		if (alive > 0) {
2191 			fi->fib_flags &= ~nh_flags;
2192 			ret++;
2193 		}
2194 
2195 		fib_rebalance(fi);
2196 	}
2197 
2198 	return ret;
2199 }
2200 
2201 #ifdef CONFIG_IP_ROUTE_MULTIPATH
fib_good_nh(const struct fib_nh * nh)2202 static bool fib_good_nh(const struct fib_nh *nh)
2203 {
2204 	int state = NUD_REACHABLE;
2205 
2206 	if (nh->fib_nh_scope == RT_SCOPE_LINK) {
2207 		struct neighbour *n;
2208 
2209 		rcu_read_lock_bh();
2210 
2211 		if (likely(nh->fib_nh_gw_family == AF_INET))
2212 			n = __ipv4_neigh_lookup_noref(nh->fib_nh_dev,
2213 						   (__force u32)nh->fib_nh_gw4);
2214 		else if (nh->fib_nh_gw_family == AF_INET6)
2215 			n = __ipv6_neigh_lookup_noref_stub(nh->fib_nh_dev,
2216 							   &nh->fib_nh_gw6);
2217 		else
2218 			n = NULL;
2219 		if (n)
2220 			state = n->nud_state;
2221 
2222 		rcu_read_unlock_bh();
2223 	}
2224 
2225 	return !!(state & NUD_VALID);
2226 }
2227 
fib_select_multipath(struct fib_result * res,int hash)2228 void fib_select_multipath(struct fib_result *res, int hash)
2229 {
2230 	struct fib_info *fi = res->fi;
2231 	struct net *net = fi->fib_net;
2232 	bool first = false;
2233 
2234 	if (unlikely(res->fi->nh)) {
2235 		nexthop_path_fib_result(res, hash);
2236 		return;
2237 	}
2238 
2239 	change_nexthops(fi) {
2240 		if (READ_ONCE(net->ipv4.sysctl_fib_multipath_use_neigh)) {
2241 			if (!fib_good_nh(nexthop_nh))
2242 				continue;
2243 			if (!first) {
2244 				res->nh_sel = nhsel;
2245 				res->nhc = &nexthop_nh->nh_common;
2246 				first = true;
2247 			}
2248 		}
2249 
2250 		if (hash > atomic_read(&nexthop_nh->fib_nh_upper_bound))
2251 			continue;
2252 
2253 		res->nh_sel = nhsel;
2254 		res->nhc = &nexthop_nh->nh_common;
2255 		return;
2256 	} endfor_nexthops(fi);
2257 }
2258 #endif
2259 
fib_select_path(struct net * net,struct fib_result * res,struct flowi4 * fl4,const struct sk_buff * skb)2260 void fib_select_path(struct net *net, struct fib_result *res,
2261 		     struct flowi4 *fl4, const struct sk_buff *skb)
2262 {
2263 	if (fl4->flowi4_oif && !(fl4->flowi4_flags & FLOWI_FLAG_SKIP_NH_OIF))
2264 		goto check_saddr;
2265 
2266 #ifdef CONFIG_IP_ROUTE_MULTIPATH
2267 	if (fib_info_num_path(res->fi) > 1) {
2268 		int h = fib_multipath_hash(net, fl4, skb, NULL);
2269 
2270 		fib_select_multipath(res, h);
2271 	}
2272 	else
2273 #endif
2274 	if (!res->prefixlen &&
2275 	    res->table->tb_num_default > 1 &&
2276 	    res->type == RTN_UNICAST)
2277 		fib_select_default(fl4, res);
2278 
2279 check_saddr:
2280 	if (!fl4->saddr)
2281 		fl4->saddr = fib_result_prefsrc(net, res);
2282 }
2283